# Model Utility Preservation on Model Utility (retain set)

Source: [Model Utility Preservation on Model Utility (retain set)](<https://www.sota2.com/research/sota/model-utility-preservation-on-model-utility-retain-set>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-05-28T06:51:18.878Z

Evaluation of model performance focused on preserving utility when applied to the Model Utility (retain set), measured by the Utility metric.

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Model Utility Preservation |
| Dataset (including any stated split) | Model Utility (retain set) |
| Metric | Utility |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 12 of 12 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [Utility](<https://www.sota2.com/research/sota/model-utility-preservation-on-model-utility-retain-set.md?metric=Utility>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Utility | Result date | Report paper |
| --- | --- | --- | --- |
| NPO | 61 | Not recorded | 68fa7c5d1e209131cea8dbff |
| UNDIAL | 61 | Not recorded | 68fa7c5d1e209131cea8dbff |
| RMU | 60 | Not recorded | 68fa7c5d1e209131cea8dbff |
| NPO | 60 | Not recorded | 68fa7c5d1e209131cea8dbff |
| UNDIAL | 60 | Not recorded | 68fa7c5d1e209131cea8dbff |
| RMU | 59 | Not recorded | 68fa7c5d1e209131cea8dbff |
| NPO | 57 | Not recorded | 68fa7c5d1e209131cea8dbff |
| NPO | 57 | Not recorded | 68fa7c5d1e209131cea8dbff |
| RMU | 57 | Not recorded | 68fa7c5d1e209131cea8dbff |
| RMU | 57 | Not recorded | 68fa7c5d1e209131cea8dbff |
| UNDIAL | 55 | Not recorded | 68fa7c5d1e209131cea8dbff |
| UNDIAL | 54 | Not recorded | 68fa7c5d1e209131cea8dbff |

### NPO

Utility: 61. Result ID: 6a17e435fbc373592c110b19.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dbff.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dbff.md>)

Recorded evaluation configuration: {"Model":"Qwen3-8B","Variant":"Context-aware"}. This record does not assert independent reproduction.

### UNDIAL

Utility: 61. Result ID: 6a17e435fbc373592c110b1d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dbff.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dbff.md>)

Recorded evaluation configuration: {"Model":"Qwen3-8B","Variant":"Context-aware"}. This record does not assert independent reproduction.

### RMU

Utility: 60. Result ID: 6a17e435fbc373592c110b14.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dbff.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dbff.md>)

Recorded evaluation configuration: {"Model":"Gemma-2B-IT","Variant":"Vanilla"}. This record does not assert independent reproduction.

### NPO

Utility: 60. Result ID: 6a17e435fbc373592c110b18.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dbff.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dbff.md>)

Recorded evaluation configuration: {"Model":"Qwen3-8B","Variant":"Vanilla"}. This record does not assert independent reproduction.

### UNDIAL

Utility: 60. Result ID: 6a17e435fbc373592c110b1c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dbff.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dbff.md>)

Recorded evaluation configuration: {"Model":"Qwen3-8B","Variant":"Vanilla"}. This record does not assert independent reproduction.

### RMU

Utility: 59. Result ID: 6a17e435fbc373592c110b1a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dbff.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dbff.md>)

Recorded evaluation configuration: {"Model":"Qwen3-8B","Variant":"Vanilla"}. This record does not assert independent reproduction.

### NPO

Utility: 57. Result ID: 6a17e435fbc373592c110b12.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dbff.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dbff.md>)

Recorded evaluation configuration: {"Model":"Gemma-2B-IT","Variant":"Vanilla"}. This record does not assert independent reproduction.

### NPO

Utility: 57. Result ID: 6a17e435fbc373592c110b13.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dbff.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dbff.md>)

Recorded evaluation configuration: {"Model":"Gemma-2B-IT","Variant":"Context-aware"}. This record does not assert independent reproduction.

### RMU

Utility: 57. Result ID: 6a17e435fbc373592c110b15.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dbff.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dbff.md>)

Recorded evaluation configuration: {"Model":"Gemma-2B-IT","Variant":"Context-aware"}. This record does not assert independent reproduction.

### RMU

Utility: 57. Result ID: 6a17e435fbc373592c110b1b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dbff.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dbff.md>)

Recorded evaluation configuration: {"Model":"Qwen3-8B","Variant":"Context-aware"}. This record does not assert independent reproduction.

### UNDIAL

Utility: 55. Result ID: 6a17e435fbc373592c110b17.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dbff.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dbff.md>)

Recorded evaluation configuration: {"Model":"Gemma-2B-IT","Variant":"Context-aware"}. This record does not assert independent reproduction.

### UNDIAL

Utility: 54. Result ID: 6a17e435fbc373592c110b16.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dbff.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dbff.md>)

Recorded evaluation configuration: {"Model":"Gemma-2B-IT","Variant":"Vanilla"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.